A position sized to lose 2% can produce a much larger realized loss when a thin order book forces the exit below the planned stop. Before a weekend, the gap between planned risk and executable risk can turn a routine trade into something closer to a 5% portfolio hit.
On August 1, 2012, Knight Capital began sending erroneous orders into U.S. equity markets. The company could not stop the problem before it had accumulated roughly $440 million in losses over about 45 minutes. Knight later required emergency financing to survive.
The U.S. Securities and Exchange Commission documented the failure in its 2013 order against Knight Capital Americas. The event was operational rather than seasonal, but its central lesson applies here: risk calculated from expected execution can become irrelevant when actual execution behaves differently.
The position looked controlled on paper
Consider Marcus, an illustrative composite trader with a $20,000 account. He reviews a stock trading near $25 on a quiet Friday in August.
His plan allows a 2% loss, or $400. He intends to buy 800 shares at $25 and exit at $24.50 if the trade fails. The calculation looks clean:
800 shares × $0.50 planned loss per share = $400.
The signal meets his entry rules. The chart shows a defined invalidation level. His journal records the planned risk. None of those facts describes what buyers may be willing to pay when he needs to sell.
That omission matters late on a thin Friday. Displayed bids can be small. Some orders may disappear as price approaches them. A stop order can trigger near $24.50 without filling there.
Suppose only part of Marcus’s position exits close to his stop. The remaining shares fill progressively lower, producing an average exit of $23.75. His loss becomes:
800 shares × $1.25 realized loss per share = $1,000.
That is 5% of the account. The position size stayed at 800 shares. The market changed what those shares could cost him.
These figures are illustrations, not a claim about any specific stock or Friday. They show why a stop price and a maximum loss are different quantities.
Thin liquidity changes the risk calculation
Position sizing usually begins with three inputs: account value, acceptable loss, and distance from entry to invalidation. That framework assumes the exit can occur reasonably close to the chosen price.
A thin order book adds another input: available liquidity between the stop and worse prices.
A visible bid for 1,000 shares does not guarantee that 1,000 shares will remain available. A stop-market order prioritizes execution over price. A stop-limit order controls price but may leave the position open. Holding through the weekend adds another uncertainty because Monday can open beyond Friday’s available prices.
The relevant question becomes: “How much can I lose if the exit fills badly?” A trader who cannot answer that should treat the position as larger than the sizing formula suggests.
Marcus could respond in several ways. He could reduce the share count, require more depth near the planned exit, avoid carrying the full position into the weekend, or reject the trade. Each choice gives up some potential participation. That cost is visible. The hidden cost appears when a $400 plan produces a $1,000 result.
The same distinction appears in The 80 Cents That Turned $25 of Planned Risk Into $65: planned risk depends on an assumed fill, while realized risk depends on the prices actually available.
Approval should test execution risk
An approval gate creates a pause between receiving a signal and placing an order. That pause has value only when the trader uses it to challenge the assumptions behind the size.
Before approving a late-Friday trade, review:
- How much size is available near the planned exit?
- What happens if the average fill is one or two price levels worse?
- Would a partial fill leave an unacceptable weekend position?
- Does the trade overlap with other positions exposed to the same market move?
- At what estimated loss does the trade become an automatic rejection?
The answer should appear in the trading journal before execution. “Risk: 2%” records an intention. “Estimated loss at normal fill: 2%; estimated loss under thin-book conditions: 4.8%” records a decision range.
If that range crosses the account’s limit, reduce the position or decline the trade. A signal can remain valid while its proposed size becomes unacceptable.
This is where approval-gated trading differs from unsupervised automation. The AI can queue a setup and show its reasoning. The human still decides whether current liquidity, concentration, and timing support the proposed exposure. That final decision should remain visible and reviewable.
Size for the exit you may receive
Knight Capital’s controls failed before the company knew the full cost of the orders entering the market. Forty-five minutes was enough to threaten the firm’s survival. The scale was extraordinary, but the practical mechanism was familiar: the system’s assumed boundaries did not contain the realized outcome.
For a retail trader, the preventive action is smaller and quieter. Recalculate the position using a stressed exit price before approving it. If a move from $24.50 to $23.75 changes the loss from tolerable to damaging, the original 800-share position was too large for those conditions.
Record both numbers. Then size from the one the market might actually deliver.
Educational content, not financial advice.
Comments
No comments yet.